Cognit ive Dissonanceafter Purchase:A M ultidim ensional Scale
J i l l i an C. S w e e n e yUniversity of Western A ustralia
Douglas Hausknecht
The U niversity of Akron
Ge o f f r e y N . So u t a rUniversity of Western A ustralia
A B S T R A C T
The concept of cognitive dissonance has been discussed widely inthe consumer behavior literature, yet paradoxically, there is nowell established scale to measure it. This article describes thedevelopment of a 22-item scale for assessing cognitive dissonanceimmediately after purchase. First, the article discusses the
multitude of studies and remains a topic of great interest in social psy-chology, as can be seen in the recent book of Harmon-Jones and Mills(1999). However, a review by Cummings and Venkatesan (1976)markeda reduction of interest in the subject in marketing, despite their conclu-sion that "the evidence in favor of dissonance theory in the consumer
behavior literature looks good" (p. 307). We agree with Oliver (1997)that this reduction in interest was both inexplicable and unfortunate.The concept needs to be further delineated and the relationships be-tween cognitive dissonance and other postpurchase constructs, such asconsumer satisfaction and attributions, need to be investigated. Thetime is ripe for a review of dissonance, a redefinition of the construct,and the development of an operational measure. Indeed, Oliver (1997,p. 261) concluded a chapter on cognitive dissonance with the hope "thatthe construction, validation and dissemination of comprehensive dis-
sonance measures will be forthcoming."Cognitive dissonance is an elusive construct. In experimental situa-
tions it has been measured in terms of indicators such as physiologicalreactions following dissonance arousal, attitude change following dis-sonance arousal or through changes in a ttitude to chosen and unchosenalternatives that were initially similarly valued. Cognitive dissonancehas been occasionally measured by one or more indicators selected byresearchers (e.g., Elliot & Devine, 1994; Menasco & Hawkins, 1978) or
by items that represent a related but different construct, such as anxiety(e.g.. Hunt, 1970).Recently, Montgomery and Barnes (1993) attempted to develop a dis-
sonance scale for use in marketing situations. Although their scale wasa meritorious attempt to clarify this elusive construct, it suffers fromseveral flaws. The various sections of the present article describe thedevelopment of a scale that attempts to measure cognitive dissonance,following Churchill's (1979) suggested procedures. First, previous mea-sures are discussed, and then the domain of the cognitive dissonance
construct is outlined. The development of the scale, including data col-lection and scale purification, is then discussed, followed by an assess-ment of the validity of the scale. The final section discusses the benefitsof having developed such a scale and its potential uses.
PREVIOUS MEASUREMENT OF DISSONANCE
Many empirical explorations of dissonance theory have manipulateddissonance experimentally within the induced/forced compliance para-
7/28/2019 11647002
http://slidepdf.com/reader/full/11647002 3/18
an issue in the opposite direction to their own opinion. In line withFestinger's dissonance theory, which specifically addresses the forcedcompliance paradigm , subjects tend to change their opinion on the issuetoward the opinion proffered in the experiment. Dissonance has beenmeasured in such cases by looking at opinions a t different stages of the
experiment. The settings of the experiments have been described asartificial, trivial, and irrelevant to marketers, because consumers arerarely trapped in a s ituation of being forced to buy a product or service(Cohen & Goldberg, 1970; Oliver, 1997; Oshikaw a, 1970).
Dissonance has also been examined through dissonance reduction.For example, Engel (1963) examined differences in the readership ofChevrolet advertisements among recent Chevrolet owners and nonown-ers, expecting that Chevrolet owners would be more likely to read theadvertisements so they could add information consonant with their pur-
chase.Kassarjian and Cohen (1965) examined the at titud es of smokers and
nonsmokers toward the believability of a report on the relationship be-tween smoking and lung cancer. Results indicated th at the more peoplesmoke, the less likely they were to believe the report, hence distortingthe cognitive elem ent (the believability of the report). Losciuto and Per-loff (1967) found that people choosing hetween two similarly attractiverecord albums subsequently ra ted the chosen one as more desirahle and
the unchosen one as less desirahle. Once again, an element (similarpreference) was distorted to be consonant with the outcome. More re-cently, Gilovich, Husted Medvec, and Chen (1995) examined the per-ceived cash value of prizes in a game show context. Subjects who hadswitched from an unopened box with a g rand prize to one with a m odestprize, placed greater cash value on their prize than participants whofailed to switch from a modest prize to a grand prize. This offers anexample of selective distortion as increasing the attractiveness of thechosen alternative decreased the attractiveness of another alternative.
When dissonance is conceptualized as dissonance arousal, other in-direct methods of establishing dissonance have been used. For example,several studies found tha t dissonance possesses arousal properties th atcan be measured in term s of task performance (Elliot & Devine, 1994).More direct measures of dissonance arousal have centered on physio-logical responses, such as galvanic sk in responses (e.g., Elkin & Leippe,1986).
Elliott and Devine (1994) argued that, although the dissonance pro-cess is initiated by arousal, the conceptualization and assessment ofdissonance should include psychological discomfort, as originally con-ceptualized by Festinger (1957). They proposed a three-item scale of
7/28/2019 11647002
http://slidepdf.com/reader/full/11647002 4/18
dissonance, using items that seemed to tap the dissonance constructwithin the context of their study. For example, Menasco and Hawkins(1978) used "difficulty of the purchase decision" to infer dissonance whenexamining the effect of various purchase conditions on dissonance. Bell(1967) used items relating to unease and rightness of the decision. Items
relating to other constructs, such as posttransaction anxiety, have alsoheen used (H unt, 1970). However, none of the studies that used ad hocmeasures attempted scale validation, and almost none (Elliot & Devine,1994, and Korgaonkar & Moschis, 1982, are exceptions) assessed scalereliability.
Montgomery and Barnes (1993, p. 206) developed a short scale ofcognitive dissonance, defining the domain of cognitive dissonance as"those feelings, attitudes and emotions that consumers have or displaywhen they experience dissonance and the situations and conditions in
which dissonance has occurred." They attempted to capture cognitivedissonance's domain through these concurrent psychological experi-ences (e.g., "dissonant consumers often display anxiety", or "dissonantconsumers may experience low levels of expected satisfaction"). How-ever, there was no framework or basis for assuming that such feelingsrepresented dissonance, beyond correlational evidence from previousstudies. In addition, they included "support" in their measure, arguing,"dissonant consumers need reassurance (e.g., support) tha t a wise pur-
chase decision has been made" (Montgomery&
Barnes, 1993, pp. 2 06 -207). The latter is a method of dissonance reduction, in contrast to dis-sonance itself. The autho rs themselves noted that this factor should bedeleted from the scale. Further, following a discussion of related re-search, the authors subjectively generated a 16-item scale, which theyfelt measured dissonance, rather than identifying dissonance issuesthrough listening to consumers who had experienced dissonance. Whilethe Montgomery and Barnes (1993) scale is a commendable attempt toclarify dissonance, the present research represents a more rigorous ap-
proach to measuring dissonance, following Churchill's (1979) scale de-velopment process.
In summary, various indirect measures have been used to establishthat dissonance occurs. Other, more direct, measures, including physi-ological measures and paper-and-pencil te sts , have also been used. Thelatter have included cognitive m easures (e.g., evaluating the purchase,the wisdom of the decision, or the actions that should have been taken)and psychological measures (e.g., anxiety, comfort, and feelings), as wellas behavioral measures designed to measure dissonance reduction.
Apart from the Montgomery and Barnes (1993) effort, useful steps inthe conceptualization-scale development sequence have not been pub-
7/28/2019 11647002
http://slidepdf.com/reader/full/11647002 5/18
THE COGNITIVE DISSONANCE DOMAIN
Festinger's early explanation of dissonance did not clearly identifywhether dissonance is cognitive or emotional. The cognitive view is sup-ported by his suggestion that "the obverse of one element follows from
the other" (Festinger, 1957, p. 261). Festinger described a person asbeing in a dissonant state if two elements in his cognition, that is, inhis knowledge of himself, his behavior, his feelings, desires, or in hisknowledge of the world, are inconsistent. Cognitive dissonance may re-sult when an opinion is formed or a decision taken when cognition andopinions d irect us in different directions. Yet Festinger (1957, p. 266)also seems to have intended an emotional conceptualization, suggestingthat, "for some people, dissonance is an extremely painful and intoler-able thing."
Cooper and Fazio (1984) considered that dissonance has less to dowith an inconsistency among cognitions per se, but rather with expec-tations of undesirable consequences. Oliver (1997) also believes disso-nance includes concern about unknown outcomes, in terms of antici-pated regret, and a feeling of apprehension on the consumer's pa rt.
Cognitive dissonance has been defined as psychological discomfort(Carlsm ith & Aronson, 1963; Elliot & Devine, 1994), a psychologicallyuncomfortable state (Festinger, 1957; Menasco & Hawkins, 1978), being
linked with anxiety, uncertainty or doubt (Menasco & Haw kins, 1978;Montgomery & Barnes, 1993; Mowen, 1995) or as synonymous with theregret or remorse reported in salespeople's anecdotes (Insko & Schopler,1972). Thus, the construct's 40-year history appears to have created atheoretical oxymoron in which a construct with an emotional themebears the burden of cognitive in its name. It is apparent that cognitivedissonance has both cognitive and psychological components.
Cooper and Fazio (1984) distinguished between the two psychologicalcomponents of "dissonance arousal" and "dissonance as a psychologi-
cally uncomfortable state." They suggested that arousal is a necessarycondition for dissonance to occur and that , if labeled negatively and anattribution is made internally (free choice, "it was my decision"), thenpsychological discomfort will arise. This psychological discomfort thenmotivates dissonance reduction, as predicted by dissonance theory. El-liot and Devine (1994) claimed that the latter element has received farless empirical attention than dissonance arousal and urged a systematicattempt to validate the psychological discomfort component of disso-nance. However, they concluded th at the distinct affect experienced byan individual in a given situation is closely related to the individual's
7/28/2019 11647002
http://slidepdf.com/reader/full/11647002 6/18
the judgmental disparity between product performance and a corre-sponding normative standard" (Westbrook & Oliver, 1991, p. 85). Hencesatisfaction, while described as emotional in nature, is based on a re-sponse to a cognitive judgment, and the construct is said to comprisecognitive as well as affective, components (Dabholkar, 1995; Oliver,
1994). Dissonance sim ilarly comprises cognitive and emotional compo-nents. As Festinger describes, it is a psychologically uncomfortablestate, but generated by inconsistent cognitions. However, there are twoimportant differences between the concepts. First, dissonance is recog-nized as immediately postdecisional (e.g., Festinger, 1957; Insko &Schopler, 1972). Satisfaction, in contrast, is assessed postpurchase andpostuse, when performance is compared to expectations. Second, satis-faction is hased on a comparison of known performance and expecta-tions, whereas dissonance concerns unknown outcomes, generating ap-
prehension that may continue after use of the product or service, whensatisfaction judgments are made (Oliver, 1997).
Oliver (1997) takes a wider view of cognitive dissonance, examiningthe concept over the entire purchase decision process. Originating in aprepurchase phase, the construct is labeled apprehension and increasesover the decision process. These same cognitions and feelings mutateinto true dissonance after the decision is made, when consideration offoregone alternatives becomes relevant. With use and experience, dis-
sonance dissipates and yields to dissatisfaction.The concept of dissonance addressed in the present research best fitsthe period that immediately follows the purchase decision but precedesuse or experience with the resu lt of the purchase decision. At this stage,labeled the "gamma" stage by Oliver (1997), dissonance is maximizedand precedes satisfaction formation.
It is important to note that dissonance is not aroused in every pur-chase. Three main conditions for such arousal have been suggested(Cummings & Venkatesan, 1976; Korgaonkar & Moschis, 1982; Mowen
1995; Oliver, 1997). First, the decision must be important to the con-sumer. That is, the consumer m ust have invested a substan tial amountof money or psychological cost in the decision and the outcome mustm atter personally to the consumer. Second, the consum er mu st feel freein making the choice. That is, the decision must be made voluntarily.Third, the consumer must display irrevocable commitment to the deci-sion once made. That is, the decision must be irreversible. Major pur-chase decisions, including those with long-term consequences, are mostlikely to create dissonance conditions (Korgaonkar & Moschis, 1982;
Oliver, 1997).Based on these definitions and distinctions, the present research
7/28/2019 11647002
http://slidepdf.com/reader/full/11647002 7/18
cision after the purchase has been made, while the emotional aspectwas defined as a person's psychological discomfort subsequent to thepurchase decision.
DEVELOPMENT OF THE SCALE
The evidence discussed suggests there are distinct cognitive and emo-tional aspects of cognitive dissonance. The presen t section describes theprocess used to establish the content of dissonance and to validate thescale psychometrically and theoretically. The process follows Churchill's(1979) approach for developing measures of multiple-item constructs.After the development of an initial set of items, two stages of scale pu-rification were undertaken. Although the first stage used two student
samples, the second stage used two m ore diverse samples of consumers.
Scale Item Generation
A total pool of over 100 dissonance items was generated from explor-atory research involving four focus groups with consumers. The proce-dure used is described at length in Hausknecht, Sweeney, Soutar, andJohnson (1998). The content validity of the items was assessed by 12
consumer behavior experts, who were provided with the cognitive andemotional definitions of dissonance used in the study. A similar proce-dure was used by Zaichowsky (1985). Item s with a significantly highermean on one dimension compared to the other were retained for furtherscale development. In all, 36 cognitive and 36 emotional items wereretained after this second stage.
One of the prime considerations in scale development is the adequacywith which a specified domain of content is sampled. The focus-groupapproach used to generate items, followed by the use of judges to assess
the items, support the scale's content validity.
Data Collection and Scale Purification, Stage 1
An initial quantitative procedure was used to reduce the number ofitems and to examine the initial scale's psychometric properties. A totalof 645 final-year students at four Australian universities participatedin this phase of the research by responding to a questionnaire containingthe to tal set of 72 items. Each student was asked to think of an impor-tant purchase decision that involved a difficult choice between two ormore close alternatives. Responses from the 455 students who had made
7/28/2019 11647002
http://slidepdf.com/reader/full/11647002 8/18
However, factor analysis has a tendency to produce too many dimen-sions for clear conceptual definition (Churchill, 1979). Coefficient betawas therefore used to clarify the existence of distinct subdimensions.Coefficient beta is a measure of reliability that considers the possibleexistence of subscales. In effect, coefficient beta is the worst split-half
reliability of a scale. Although coefficient alpha rep resents the averageof all split-half reliabilities (Cronbach, 1951) it does not reveal a lowsplit-half reliability that can indicate the presence of a subscale (John& Roedder, 1981). The approach for estimating coefficient beta is dis-cussed by John and Roedder (1981) and Revelle (1979).
The results of the coefficient beta analysis for the 72-item scale sug-gested a single emotional dimension and three cognitive dimensions.Many emotional items (e.g., angry, annoyed, frus trated, depressed) rep-resent the negative end of the pleasure dimension in the psychologicalspace of consumption emotion discussed by Bush (1973) and Russell(1980). Some items also represent the higher end of Bush's aggressiondimension (angry, furious with myself), as well as the h igher end of theBush and Russell arousal dimensions (excited, angry, annoyed). Inter-estingly, if one considers Plutchik's (1980) circumplex of emotions, inwhich eight basic emotions ranging from joy to anger are arranged in aring, the emotional items in this study can be found in a specific segm entrepresen ting a th ird of the circumplex area. This suggests some restric-
tions to the negative emotions th at relate to dissonance.Cognitive items, in contrast, rela te to though ts about the wisdom ofthe purchase decision that took place. The majority relate to self-attri-bution (e.g., "I thought I shouldn't have done it," "I wondered if I couldhave made a better buy"), although a few concerned the favorability ofthe deal obtained in the store (e.g., "I wondered if they were spinningme a line"). The three dimensions that emerged related to whether theright choice was made, whether the product was really needed at all,and concern over the deal, particularly in regard to the salesperson.
Coefficient beta values were .81 or above for the four scales, and asso-ciated conditions recommended by John and Roedder (1981) were sat-isfied, suggesting tha t they rep resent unidimensional aspects.
Scale content and reliability were examined and, as recomm ended byChurchill (1979), the scale was further reduced by plotting item to thetota l scale correlations for each factor. Item s th at produced a sharp dropin the plotted pattern were eliminated. Coefficient alpha was then re-calculated for the remaining items and the procedure repeated untilthere were no significant drops in the correlations. Scales were severelytrimm ed at th is stage, since the intention was to develop a parsimonious
7/28/2019 11647002
http://slidepdf.com/reader/full/11647002 9/18
Table 1. Confirmatory Factor Analysis of Models of Various
Dimensionalities.
Model
Stage 1
Null
One factorTwo factors (cognitive and emotional)Four factors (as proposed in Stage 1)
Three factors (solution after Stage 2 on
reduced set of Stage 1 items)
Stage 2
Furniture storeNull
One factor
Three factors (as proposed in Stage 2)Car stereo store
NullOne factorThree factors (as proposed in Stage 2)
6597.64
2197.601259.16
453.53288.36
3996.73915.72
346.52
3502.83850.35368.17
df
136
119118
113
5 1
66
54
5 1
66
5 4
5 1
RM R
.4 6
.1 5
.11
.05
.09
.5 0
.1 4
.05
.50
.14
.0 7
RNI
.6 8
.8 2
.95
.9 5
.7 8
.92
.7 7
.91
' The root mean-square residual is denoted by RMR and the relative noncentrality index by RNI. RNI =
' - dfj - (.x" - df)\Hx^^ - dfj, where n is the null model.
ues df coefficient alpha for the four subscales following this processranged from .85 to .96. An exploratory factor analysis of these itemssupported the four-factor structure. Confirmatory factor analysis,through structural equation modeling, indicated that the four-factormodel was superior to a null model, a single-factor model in which allitems represented a single dissonance factor and a two-factor model, inwhich emotional and cognitive elements represented two separate di-mensions (Table 1). This supports the dimensionality and the discrim-inant validity of the suggested scale dimensions.
The discriminant validity of the four-dimensional scale was also in-vestigated in two further ways. First, the test that the correlation be-tween constructs is significantly less than one was used (Bagozzi &Heatherton, 1994). As can be seen in Table 2, the highest correlationbetween dimensions was .62 and the associated confidence in terval was.56 to .68. This test supported the discriminant validity for all pairs ofdimensions. Second, the Fornell and Larcker (1981) discrim inant valid-ity tests were conducted. These tests require th at the average varianceextracted be greater than .50 and, when taking any pair of constructs,
that the average variance extracted for each construct is greater thanthe squared structural path coefficient between the two constructs. In
7/28/2019 11647002
http://slidepdf.com/reader/full/11647002 10/18
Table 2. Correlations Between Constructs following Confirmatory Factor
Analysis.
Emotional
Made RightChoice? Did I Need It?
Concern Over
Deal
Stage 1
EmotionalM ade rig ht choice?Did I need it?
Concern over deal
.97
.50 (.04)
.65 (.03)
.62 (.03)
.91
.53 (.04)
.62 (.03)
.82
.54 (.04) .86
EmotionalWisdom ofPurchase Concern Over Deal
Stage 2Furniture store
EmotionalWisdom of purchaseConcern over deal
Car stereo store
EmotionalWisdom of purchase
Concern over deal
.98.52 (.05)
.69 (.03)
.96
.47 (.05)
.74 (.03)
.81
.43 (.03)
.83
.53 (.05)
.86
.80
Note: Reliability of linear com posite of scale Stage 1 (28 items) = 0.98.Reliability of linear composite of scale Stage 2 (22 items) furnitu re store = .96, car stereo store .97. Stand ard
errors appear in brackets, while composite reliabilities appear on the diagonal.
inant validity of the four dimensions, even when measurement errorwas considered.
The entire analysis was repeated on a smaller but similar studentsample {n = 183). The analysis (not shown) revealed similar results.However, the 31-item scale was further reduced to 28 item s. Two of thethree items deleted were found to have lower item to total correlations
than other items representing that dimension and one showed overlapacross the dissonance dimensions. The meaning of these subscales wasunaffected by the omission of these items and, hence, for the sake ofparsimony, the three items were deleted.
Although these results provided evidence of construct validity, theywere based on a student sample. The 28-item ins trum ent was thereforereexamined with the use of an independent and more diverse seconddata set, as recommended by Churchill (1979).
Data Collection and Scale Purification, Stage 2
7/28/2019 11647002
http://slidepdf.com/reader/full/11647002 11/18
selling durable goods. These were a furniture store (two outlets) and acar stereo center (three outlets). Customers were invited to participatein the survey imm ediately after comm itting themselves to the purchase.Only customers making a major purchase (defined as those spendingover $400 on the purchase) were included in the sample. Such customers
were asked to take a self-completion questionnaire and return it in thereply paid envelope within 10 days. An opportunity of participating ina draw for a voucher redeemable at the store was used as an incentiveto boost response. The survey was conducted over a 6-month time periodand all qualified customers were asked to participate. Usable responsesrepresented 44% of furniture store customers and 31% of car stereo cen-ter customers.
To check for non-response bias, the sample profiles in te rm s of basicvariables, such as suburb of residence and expenditure levels, were com-pared with company records of all such customers over the same timeperiod. A chi-square test found no significant differences. It was con-cluded that the sample was representative of major purchasers duringthis time period.
Because it was dissonance in the postpurchase-pre-use stage thatwas of interest in the p resent study, furniture store customers who hadnot yet taken delivery of the ir purchases were selected for analysis . Mostrespondents were in this category, because furniture would typically
take between 1 and 6 weeks to deliver. However, in the car stereo center,equipment was typically installed immediately. Almost 70% of respon-dents completed their survey after installation and experience with theequipment. The car stereo sam ple was therefore restricted to those whocompleted the questionnaire within 10 days of installation. Final samplesizes for analysis were 224 for the furniture store and 313 for the carstereo center. The furniture store sample was taken as the primary sam-ple for analysis due to the more rigorous natu re of the sample (the longertime period and hence ability to complete the questionnaire before de-
livery of the product). The analytical section that follows, therefore, isbased largely on the furniture store sample.
Analyses
As in the first stage, items that created a sudden drop in the plotteditem to total scale correlation pa ttern s w ere dropped. Three items weredeleted as a result of this process. Alpha coefficients as a resu lt of thisranged from .80 to .97 (furniture) and from .78 to .95 (car stereo).
7/28/2019 11647002
http://slidepdf.com/reader/full/11647002 12/18
monality, because both concerned internal attributions, in contrast tothe other cognitive dimension "concern over the deal," which supposessome degree of external a ttribution in the role of the salesperson. Thesetwo dimensions were therefore combined to form a "wisdom of purchase"subscale. Based on this three-factor model, three items in the explor-
atory factor analysis that had split loadings were dropped.A further exploratory factor analysis supported the three-dimen-
sional solution. Once again confirmatory factor analysis and reliabilityestimates were used to evaluate the validity and reliability of the re-duced scale. Confirmatory factor analysis supported the three-dimen-sional solution (Table 1), compared to the null, one-, and two-factor mod-els for both samples. Correlations between the dimensions weresignificantly less than one (Table 2), further supporting the discrimi-
na nt validity of the dimensions. Additionally, the variance extracted foreach dimension was .52 or greater which, in each case, exceeded thesquare of the path between any pair of constructs, which had a maxi-mum of .49.
Due to the different na tures of the samples in Stages 1 and 2 and thechange in the dimensional structure adopted at the beginning of thesecond stage (the single factor "wisdom of purchase" replacing the twoprevious factors "whether the right choice was made" and "whether theproduct was really needed at all"), the confirmatory analysis and tests
of discriminant validity for the three-factor model were also conductedon the original student sample. The fit was extremely good, being su-perior to the original fit of the 31 items on four dimensions and also tothe fina l fit on both second stage data sets (Table 1).
In summary, there was strong support for the discriminant validityof the suggested th ree cognitive dissonance dimensions. The dimensionswere defined as follows:
Em otional A person's psychological discomfort sub seq uen t to the purc has edecision
Wisdom of purchase A person's recognition after the purchase has heen made th atthey m ay not have needed the product or may not have se-letted the appropriate one
Concern over deal A person 's recognition after the purc has e ha s heen ma de th atthey may have heen influenced against their own beliefs hysales staff
Specific details of the final 22 items used in the scale are shown in
Table 3. The criterion-related validity of the scale (i.e., whether the m ea-sure behaves as expected with measures of other constructs external to
7/28/2019 11647002
http://slidepdf.com/reader/full/11647002 13/18
Table 3. Fina l Scale (22 Items).
Emotional
Wisdom of purchase
Concern over deal
After I bought this product:I was in despairI resented it
I felt disappointed with myselfI felt scaredI felt hollowI felt angryI felt uneasyI felt I'd let myself downI felt annoyedI felt frustratedI was in painI felt depressedI felt furious with myselfI felt sick
I was in agony
I wonder if I really need this productI wonder whether I should have bought anything at allI wonder if I have made the right choiceI wonder if I have done the right thing in buying this
product
After I bought this product I wondered if Fd been fooledAfter I bought this product I wondered if they had spun me
a line
After I bought this product I wondered wh ether the re wassomething wrong with the deal I got
experience satisfaction, and would have experienced more difficulty inassessing the quality of the product. For the present purposes, the sam-ples were restricted to respondents who had not received the ir furniturepurchase or not had their car stereo equipment installed (i.e., those who
were in the gamma phase of the purchase decision process). The rele-van t correlations are shown in Tahle 4.
The expectation that high-dissonance consumers would have hadgreater difficulty in judging the quality of the product and that disso-nance would result in lower levels of satisfaction and value were sup-ported in hoth samples. The relationship was strongest with qualityjudgment difficulty, although the association of dissonance with satis-faction was suhstantial. The relationship was weakest, hut still signif-icant, w ith perceived value. As might have heen expected, quality judg-
ment difficulty was most related to people's "concern over the deal,"because difficulty in judging quality could he allayed hy salespeople's
7/28/2019 11647002
http://slidepdf.com/reader/full/11647002 14/18
Table 4. Relationship of Dissonance Dimensions and Related Constructs,
Correlational Analysis. Product Not Yet Delivered or Installed (Gamma
Phase).
Related Construct
Furniture store
I will value this product
I feel satisfied with my decision to huythis product
I had considerable difficulty in judg-ing the quality of the product
Car stereo storeI will value this product
I feel satisfied with mydecision to buythis product
I had considerable difficulty in judg-ing the quality of the product
Dimension
Concern over dealEmotionalWisdom of purchase
Concern over deal
EmotionalWisdom of purchase
Concern over dealEmotionalWisdom of purchase
Dimension
Concern over dealEmotionalWisdom of purchase
Concern over dealEmotionalWisdom of purchase
Concern over dealEmotionalWisdom of purchase
(n = 224)
- . 3 0 * *- . 1 7 *- . 1 4 *
- . 3 5 * *- . 3 1 * *
- . 1 7 *
.51**
.42**
.27**
(n = 94)
- . 4 8 * *- . 3 3 * *- . 1 2
- . 4 9 * *- . 4 5 * *- . 1 5
.64**
.37**
.23*
*p < .01.
**p < .05.
mensions, although the emotional component of dissonance also playeda significant role here. The cognitive component, "wisdom of purchase"
had a weaker relationship with the three related constructs, but wasmost strongly related to difficulty in judging the quality of the good.Overall, the criterion-related validity of the scale was supported.
DISCUSSION
Despite considerable interest in the concept of cognitive dissonance,there has only been one attempt at developing a scale for the construct(Montgomery & Barnes, 1993). This scale has its drawbacks, however,
7/28/2019 11647002
http://slidepdf.com/reader/full/11647002 15/18
use stage of the decision process, duhhed hy Oliver (1997) as the gammaphase. The study suggests that dissonance includes hoth cognitive as-pects, as the title cognitive dissonance implies, as well as an emotionaldimension, as many definitions, including Festinger's original defini-tion, imply. The 22-item scale that w as developed tapped three dimen-
sions of dissonance, including one emotional and two cognitive dimen-sions.
The study was hased on four samples. Two student samples were usedin Stage 1 to test and initially reduce the developed set of items, whiletwo customer samples were used in Stage 2 to test the reduced set onreal consumers who were at the gamma phase of the decision process.The developed scale had strong content, discriminant, and Criterion-related validity, as well as a high level of reliahility, and should provean asset in consumer hehavior research. For example, Elliott and De-vine (1994) have urged researchers to explore the inner workings of hothaffective and cognitive processes in the dissonance induction-reductionsequence. Further, Oliver (1997) has suggested that dissonance induc-tion and dissonance reduction operate in a dynamic model in the hetaand gamma stages of the decision-making process (postdecision hu t pre-use) and that, depending on their relative strengths, various outcomes,such as regret cognitions and disconfirmation, are initiated and affectcustomer satisfaction.
Dissonance is known to follow a personal decision hut, as indicated,may continue over the entire decision-making process and may nevercompletely disappear (Oliver, 1997). The presence, magnitude, and ef-fects of dissonance need to he studied over all stages of the decision-making process, including at the time of repeating the purchase. Thedeveloped scale will assist in these explorations.
REFERENCES
Bagozzi, R., & Heatherton, T. F. (1994). A general approach to representingmultifaceted personality constructs: Application to state self-esteem. Struc-tural Equation Modeling, 1, 35-67.
Bell, G. D. (1967). The autom obile buyer after pu rch ase . Jo ur na l of M ark eting ,31, 1 2 - 1 6 .
Bush, L. E., II (1973). Individual differences multidimensional scaling of ad-jectives denoting feelings. Journal of Personality and Social Psychology, 25,5 0 - 5 7 .
7/28/2019 11647002
http://slidepdf.com/reader/full/11647002 16/18
Cohen, J. B., & Goldberg, M. E. (1970). The dissonan ce model in post-decisionproduct evaluation. Journal of Marketing Research, 7, 3 1 5 - 3 2 1 .
Cooper, J., & Fazio, R. H. (1984). A new look a t d issonance theory . In L. Berkowitz (Ed.), Advances in ex perim ental social psychology (Vol. 17, pp. 2 2 9 -266). San Diego, CA: Academic P res s.
Cronbach, L. J. (1951). Coefficient alpha and the internal structure of testsPsychometrika, 16, 297-334.
Cum mings, W. H., & Venk atesan , M . (1976). Cognitive dissonance an d consumer behavior: a review of the evidence. Journal of Marketing Research13 , 3 0 3 - 3 0 8 .
Dabholkar, P. (1995). A contingency framework for predicting causality be. tween custom er satisfaction an d service qua lity. Advances in Co nsum er Re
search, 22, 101-108.Elkin, R., & Leippe, M. (1986). Physiological arousal, dissonance and attitude
change: evidence for a dissonance-arousal link and a "don't remind me" ef-
fect. Jou rna l of Person ality and Social Psychology, 51, 5 5 -6 5 .Elliot, A. J., & Devine, P. G. (1994). On the motivational nature of cognitive
dissonance: dissonance as psychological discomfort. Journal of Personalityand Social Psychology, 67, 3 82 -3 94 .
Engel, J. F. (1963). Are automobile purchasers dissonant consumers? Journalof Market ing, 27, 55 -5 8 .
Fes tinge r, L. (1957). A theo ry of cognitive dissona nce. Stanford , CA: StanfordUniversi ty P ress.
Fornell, C , & Larcker, D. F. (1981). Ev alua ting s truc tura l equation m odels
with unobservable variables and m^easurement error. Journal of MarketingResearch, 18, 39-50.
Gilovich, T., Hu sted Medvec, V., & Chen, S. (1995). Com mission, omission, a nddissonance reduction: coping with regret in tbe "Monty Hall" problem. Per-sonality and Social Psychology Bulletin, 21 , 182 -1 90 .
Harmon-Jones, E., & Mills, J. (1999). Cognitive dissonance: progress on a piv-otal theory in social psychology. Was hing ton, D C: Am erican PsychologicaAssociation.
H aus kn ech t, D., Sweeney, J. C , Sou tar, G. N., & Joh nso n, L. W. (1998). "After
I had m ade tbe decision, I . . . :": Toward a scale to m eas ure cognitive dis-sonance. Jo urn al of Consum er Satisfaction, Dissatisfaction and Com plainingBehavior, 11, 11 9-1 27 .
H un t, S. D. (1970). Post-transa ction comm unications an d dissonance redu ctionJou rnal of M arketing, 34, 4 6 - 5 1 .
Insko , C. A., & Schopler, J. (1972). Ex pe rim enta l social psychology: Text w ithillustrative reading s. New York: Academic Pre ss.
John, G., & Roedder, D. L. (1981). Reliability assessment: Coefficients alphaand beta. In K. Bernhardt et al. (Eds.), Tbe changing marketing environ-ment: New theories and applications (pp. 154-157). Chicago, IL: AmericanMarketing Association.
7/28/2019 11647002
http://slidepdf.com/reader/full/11647002 17/18
Losciuto, L. A., & Perloff, R. (1967). Influence of product preference on disso-nance reduction. Jou rnal of M arketing Research, 4, 286 -29 0.
Menasco, M. B.,& H aw kin s, D. 1. (1978). Afield t est of the re lation ship betw eencognitive dissonance and state anxiety. Journal of Marketing Research, 15,650 - 655 .
Montgomery, C, & Barnes, J. H. (1993). POSTDIS: A short rating scale form easu ring postpu rchase dissonance. Jou rna l of Consum er Satisfaction, Dis-satisfaction and Complaining Behavior, 6, 204-216.
Mow en, J. C. (1995). Co nsum er beh avior (4th ed.), Englewood Cliffs, NJ : Pre n-tice Hall.
Oliver, R. L. (1994). Conceptual issues in the structural analysis of consump-tion emotion, satisfaction and qua lity: Evidence in a service setting . In C. T.Allen & D. R. Jo hn (Eds.) , Advances in co nsumer research (Vol. 21 , pp. 1 6 -22). Association for Consumer Research.
Oliver, R. L. (1997). Satisfaction: A behavioral perspective on the consumer.New York: McGraw-Hill.
Oshikawa, S. (1970). Consumer pre-decisional conflict and post-decision dis-sonance. Behavioral Science, 15, 132 -1 40 .
Plutchik, R. (1980) Emotion: A Psychoevolutionary synthesis. New York: Har-per and Row.
Revelle, W. (1979). Hierarchical cluster ana lysis an d th e in tern al s truc ture oftests . M ultivar iate Behavioral Research, 14, 5 7- 7 4 .
Ru ssell, J. A. (1980). A circump lex model of affect. Jo ur na l of Pers on ality andSocial Psychology, 39, 1161-1168.
Westbrook, R. A., & Oliver, R. L. (1991). The dimensionality of consumptionemotion pa tter ns and consumer satisfaction. Jour na l of Consum er Research,18 , 84-91 .
Zaichowsky, J. L. (1985). M easu ring th e involve men t construc t. Jou rn al of Con-sumer Research, 12, 341-352.
The authors thank Lester Johnson of Monash University, Melbourne, for hisassistance in the data collection in Stage 1.
Correspondence regarding this article should be sent to: Jillian C. Sweeney,
University of Western Australia, Perth, WA 6009, Australia (jsweeney®-ecel.uwa.edu.au).
